Pith. sign in

REVIEW 3 major objections 7 minor 1 cited by

Likelihood-free Model Selection in Cosmic Reionization with Three-dimensional Tomographic 21 cm Lightcone Images

T0 review · 3 major / 7 minor · reviewed 2026-08-08 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Simulation-based Bayesian evidence picks the correct reionisation model from a 21 cm lightcone in every mock test.

desk verdict An honest methods paper that builds a new likelihood-free model selection pipeline for 21 cm reionization and documents a real failure mode, but the Bayes factor calibration rests on an assumed proportionality constant that the paper's own posteriors suggest is model-dependent. read the letter →

arxiv 2502.08152 v1 pith:RNK6DT3L submitted 2025-02-12 astro-ph.IM astro-ph.CO

T. Binnie , Xiaosheng Zhao , J. R. Pritchard , Yi Mao This is my paper · ORCID
classification astro-ph.IMastro-ph.CO
keywords cosmicreionization21cmcosmologylikelihood-freeinferenceBayesianmodelselectionBayesfactorconvolutionalneuralnetworknestedsamplingtomographiclightcone
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that Bayesian model selection for the epoch of reionisation can be done without writing down a likelihood: a convolutional network compresses each 21 cm lightcone into summaries, a neural density estimator learns the joint density of parameters and summaries, and nested sampling integrates that learned density over the prior to give an evidence value. If this works, 21 cm data collected by upcoming interferometers can be compared between competing reionisation scenarios directly from simulations, sidestepping analytic likelihoods that are hard to write down. The paper reports that the resulting Bayes factors identify the correct model in every one of the four mock data sets, with strong odds, and the runner-up model always shares the mock data's inside-out or outside-in morphology. It also reports that the same pipeline recovers fiducial parameters for inside-out models, while outside-in models miss them by 3-6 sigma, signalling that the learned density does not always map correctly onto the posterior. The paper nevertheless concludes that the evidence values remain usable for model comparison, and urges scrutiny of how flexibly CNN-based summaries transfer to models outside their training conditions.

What carries the argument

The load-bearing object is the learned joint density $\rho(\theta, D)$: the posterior is taken proportional to $\rho$ evaluated at the mock data, and the evidence is its integral over the prior. The pipeline compresses each lightcone with a 3D convolutional network into summaries, applies score-based compression, fits the joint density with neural density estimators, and then replaces Markov-chain sampling with nested sampling to perform the evidence integral. The Bayes factor is the ratio of two such integrals; the argument only works if the unknown proportionality constant between $\rho$ and the true posterior cancels in the ratio.

What would settle it

Generate mock 21 cm lightcones from an outside-in model at several fiducial points and compare the log-evidence differences produced by the learned-density pipeline with analytic-likelihood evidence computed from the same summaries; if the evidence gap moves with the fiducial parameters, the proportionality constant is parameter-dependent and the Bayes factors are not valid.

Watch

Extended reading notes

Core claim

The authors construct a likelihood-free model-selection pipeline in which a convolutional network compresses three-dimensional 21 cm lightcones into summaries, a neural density estimator learns the joint density of parameters and summaries, and nested sampling integrates that learned density over the prior to deliver an evidence value. They claim this is the first likelihood-free Bayesian model selection in reionisation, and report that the Bayes factor favours the true generating model with strong odds for all four toy morphologies, with the runner-up always sharing the mock data's inside-out or outside-in morphology. They also report that inside-out models recover their fiducial parameters, whereas the two outside-in models miss the fiducial parameters by 3-6 sigma, which they attribute to model- and parameter-dependent information loss in the CNN compression. The paper's conclusion is that the evidence values remain usable for model comparison, but the flexibility of CNN-based summaries across morphologies needs scrutiny before application to real data.

Load-bearing premise

The whole comparison trusts that the neural network's learned density is off by the same factor for every model, so that factor cancels when the two models are divided.

Editorial extensions

If this is right

  • If the proportionality constant cancels, Bayes factors from the pipeline are usable model-selection odds without writing an analytic likelihood for the 21 cm signal.
  • For the four toy morphologies tested, the correct model wins with odds ranging from hundreds to one to about 10^13 to one.
  • Because the runner-up always shares the mock model's morphology, the model ranking appears to track physical morphology rather than accidental summary similarity.
  • The 3-6 sigma posterior failures on outside-in models imply that a confirmed model ranking does not guarantee an accurate parameter posterior.
  • A larger 250 Mpc simulation box broadens posteriors relative to the smaller 66 Mpc box used in the original CNN work, so future training data must include realistic cosmic-variance scales.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If real reionisation is a mix of inside-out and outside-in phases, the reported outside-in posterior failure implies that using these summaries on observed data could bias parameter inference even when model ranking looks decisive.
  • A direct calibration check would compare the learned evidence against an analytic-likelihood evidence for a two-model problem with known normalisation, repeated at several fiducial points; a shift in the evidence ratio with the fiducial point would falsify the proportionality-cancellation assumption.
  • Summaries that explicitly encode void and bubble edge statistics, or information-maximising compressions, might extend the method's comfort zone; this is testable by retraining the same density-estimation and integration pipeline with those summaries.
  • The strong odds in the four-model test may partly reflect the extreme morphological contrast between toy models; distinguishing models whose lightcones differ more subtly would test whether the reported rankings survive realistic discrimination.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 7 minor

Summary. This paper develops a likelihood-free (simulation-based) model-selection pipeline for epoch-of-reionization (EoR) morphology studies. A 3D CNN from Zhao et al. (2022a) compresses 21 cm lightcone images into two summary statistics; pyDelfi learns a joint density over parameters and summaries; and the Emcee sampler is replaced with MultiNest so that the learned joint density can be integrated to produce an approximate Bayesian evidence. The pipeline is tested on mock data from four toy EoR models (FZH, FZHinv, MHR, MHRinv) built on 21cmFAST with contrasting inside-out and outside-in morphologies. The paper reports that in all four cases the correct generating model receives the largest evidence with strong odds, with the runner-up model generally sharing the morphology of the true model. It also reports, and discusses at length, that the parameter posteriors for the two outside-in models fail to recover the fiducial parameters, while the inside-out posteriors succeed. Additional cross-checks include a cosmic-shear validation of the sampler swap, importance nested sampling, a 2D-CNN variant, a no-recurring-seeds simulation variant, and +/-10% prior-range variations. The Delfi-Nest code is released publicly.

Significance. The paper makes a genuine methodological contribution and is, to my knowledge, the first application of likelihood-free model selection to EoR morphology. Its strengths are the breadth and honesty of the validation program: the cosmic-shear comparison (Section 6.1) shows that substituting MultiNest for Emcee integrates the learned joint density in agreement with the original pyDelfi pipeline; the importance-nested-sampling comparison (Appendix D.4) checks sampler convergence; the 2D-CNN, no-recurring-seed, and prior-variation experiments (Appendices B.1, B.2, B.4) test the robustness of the model ranking; and the public release of Delfi-Nest makes the machinery reproducible. The paper also reports plainly that the outside-in posteriors fail to recover the mock data parameters (Table 4, Figure 6), an outcome that many papers would omit.

major comments (3)
  1. [Section 6, Eqs. (7)-(8), Table 4] The central claim that the integrated joint density yields usable Bayes factors rests on the assumption, stated in Section 6, that "any constant of proportionality between the joint density and Posterior obtained in Equation 7 ... will be largely cancelled." This assumption is not validated and is contradicted by the paper's own results: the MHR|MHR posterior (Table 4) recovers log10Tvir = 5.39 +/- 0.07 against the mock fiducial 5.80 (a roughly 6 sigma offset), and FZHinv|FZHinv recovers zeta = 2120 +/- 1290 against the fiducial 72. When the slice rho(theta, D_f) is sharply peaked in the wrong region of parameter space, the integral Z_hat in Eq. (8) is not the marginal likelihood of the data under the model, so the Bayes factors in Figure 5 are not calibrated odds. The cosmic-shear cross-check (Sections 6.1-6.2) cannot resolve this because the two shear models share the same score-compression function and an approximately Gaussian likelihood. The extreme values in Table 4 (lnZ(FZHinv|FZHinv) = 17.3 despite missing the fiducial zeta by more than an order of magnitude in a prior spanning three decades; off-diagonal lnZ values as low as -453) are hard to reconcile with genuine evidence integrals. I recommend either reframing the results as a summary-space discriminant that is explicitly not a calibrated probability, or adding a calibration test, for example computing analytic-likelihood evidences with the 21 cm power-spectrum pipeline of Binnie & Pritchard (2019) on the same mock data and quantifying the offset from the Delfi-Nest evidences.
  2. [Section 3 and Eqs. (7), (15)] The Bayes factor in Eq. (15) compares evidences obtained by integrating joint densities built from different, model-specific summaries: each model has its own 3D CNN, trained on that model's lightcones, so the computed ratio is B_ij = integral rho_i(theta, t_i(D_f)) dtheta / integral rho_j(theta, t_j(D_f)) dtheta with t_i, which is not equal to t_j. The ratio of marginal densities of two different summary statistics at the same data set equals the intended evidence ratio only if each summary is sufficient for its model, or if the summary functions are consistent across models; neither is established, and the failure of the outside-in posteriors to recover their fiducial parameters is direct evidence that the CNN summaries are not sufficient for those models. The cosmic-shear cross-check does not bear on this point because both shear analyses use the same score-compression statistic. The paper should state this comparability problem explicitly and, as a diagnostic, compare against a fixed summary statistic applied to all four models.
  3. [Sections 7.2, 8, 9] The validation presented for the model-selection numbers is at the level of ranking stability rather than evidence calibration. The agreement with Binnie & Pritchard (2019) invoked in Section 9 is a ranking agreement only: the earlier likelihood-based analysis recovered the fiducial parameters for all four models, whereas the present pipeline fails to do so for the two outside-in models (Table 4), so agreement on the winning model does not validate the evidence scale. The alternate analyses (2D-CNN, no-recurring-seed, +/-10% priors, importance nested sampling) all preserve the ranking, which is a useful robustness statement, but a ranking-preserving transformation can change the odds by orders of magnitude. Because the mock data are generated in-domain by the same 21cmFAST pipeline used for training, the strong odds (10^6 to 10^8) measure in-distribution separability of learned summaries; the abstract and conclusions should state this closed-box scope explicitly next to the "Bayes factor" and "betting odds" terminology introduced in Section 5.1.
minor comments (7)
  1. [Section 7.2] The text states that for FZHinv|FZHinv "zeta is ~3 sigma too large despite a variance of 1290"; with the quoted uncertainty 1290 the offset is about 1.6 sigma, and the use of "variance" for what appears to be a standard deviation is ambiguous. In the same paragraph, the MHR|MHR log10Tvir MAP of 5.39 is said to be "nearly 6 sigma higher" than the fiducial 5.80, but it is lower.
  2. [Sections 4.1 and 8] The "pseudo-posterior" and "pseudo-Evidence" terminology is introduced in Section 4.1 and then dropped on the grounds that "these should converge"; since Section 8 concludes that the joint density does not correctly map to the posterior for outside-in models, the cautious terminology (or an explicit pointer to where convergence fails) should be retained where the hat notation is abandoned.
  3. [Section 9] The code name is given as "pyDelphi" in the Conclusions and "pyDelfi" elsewhere; please make the spelling consistent throughout.
  4. [Section 3 and Appendix B.4] The claim that the summary space is "unchanged to <1%" under +/-10% prior changes is used to argue that the Evidence is unchanged, but the Evidence integral is taken over the parameter prior; a +/-10% prior change should move lnZ by of order 0.1-0.2 per dimension, so the two statements should be reconciled explicitly.
  5. [Section 5.2] The dismissal of Dynesty as increasing "computational intensity with negligible benefit" is given without quantitative support; a one-sentence justification (wall-clock time or numbers of likelihood evaluations) would make this reproducible.
  6. [Figures 4 and 6] The x-axis is labelled "Log10[Mvir]" while the text and Table 4 quote log10Tvir for the same numerical values; since Tvir is proportional to Mvir^(2/3), please clarify whether the axis is mislabeled or a conversion is being applied.
  7. [Figure 5] The morphological coding by red/blue color plus grey-scale Jeffreys' bands will be difficult to read in grayscale print; please add hatched patterns, symbols, or line styles.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the Delfi-Nest evidence calculation is an approximate simulation-based estimator whose proportionality assumption is explicitly flagged, and the model-selection test is an in-distribution validation rather than a derivation forced by construction.

full rationale

The paper's central derivation is the likelihood-free evidence estimate: pyDelfi learns a joint density rho(theta,D) per model, and Delfi-Nest integrates it with MultiNest to form a pseudo-evidence (Eqs. 7-8), whose ratios are reported as Bayes factors (Eq. 15). This is not circular by construction: the pseudo-evidence is an integral of a fitted density, and the paper explicitly distinguishes it from the likelihood-based evidence, flags the proportionality assumption in Section 6 ('We assume that any constant of proportionality between the joint density and Posterior obtained in Equation 7 ... will be largely cancelled'), and candidly reports that outside-in models produce posteriors that miss the fiducial parameters (Table 4, Fig. 6), which is a calibration failure rather than a definitional identity. The model-selection test is an in-distribution validation: mock lightcones are drawn from the same four 21cmFAST toy models used to generate training data, so success shows the summaries are informative, but it does not reduce the claimed Bayes factors to their inputs. The self-citations (Z22 CNN, Binnie & Pritchard 2019 models and previous model-selection setup) are backed by independent reproductions within the paper (Section 7.1 reproduces Z22 posteriors; Section 6.1 reproduces A19 posteriors), so they are not load-bearing. The strongest validity concern, that the unknown proportionality constant may be model-dependent, is explicitly acknowledged and is a matter of calibration rather than circularity.

Assumptions & free parameters 3 free parameters · 6 assumptions · 2 invented entities

The central methodological claim rests on standard statistical machinery, the 21cmFAST simulation assumption, and two ad hoc assumptions about the CNN summaries and the proportionality constant. The four toy models themselves are ad hoc test beds.

free parameters (3)
  • Prior ranges for zeta and log10Tvir per model = FZH: zeta in [10,250], log10Tvir in [4,6]; FZHInv: [10,5000], [4,7]; MHR: [10,4000], [4,7]; MHRinv: [10,1200], [4,6]
    These uniform priors are taken from Binnie & Pritchard (2019). They set the training-data distribution for the CNN and the integration volume for the Evidence; the Bayes factors are conditional on them, though the paper shows 10% changes do not change results.
  • Mock data fiducial parameters = FZH: (30, 4.7); FZHInv: (72, 4.72); MHR: (413, 5.80); MHRinv: (268, 5.61)
    Single chosen parameter points define each mock data set. The success of posterior recovery is tested at these points only; the failure for outside-in models may depend on the chosen fiducial values.
  • Number of training lightcones per model = 10000
    Chosen by the authors for CNN training; a larger number might improve summaries for outside-in models, as the authors speculate in Section 8.
assumptions (6)
  • standard math Bayes theorem, nested sampling (MultiNest), and the use of the learnt joint density as a surrogate posterior (Eqs. 7, 8, 14-16)
    The statistical framework is standard, but the validity of integrating an NDE-based joint density to obtain Evidence requires that the proportionality in Eq. (7) is exact or cancels.
  • domain assumption 21cmFAST lightcones are valid mock observations of the 21 cm signal for the redshift range z ~ 7.5-12
    All training and mock data are drawn from this semi-numerical simulator. The paper does not test against hydrodynamic simulations or real data; the lightcone construction repeats structure every delta z = 1.03.
  • ad hoc to paper The 3D CNN summaries t(theta) are sufficient statistics for (zeta, log10Tvir) for all four toy models
    The CNN is trained separately for each model and its summaries feed the density estimator. The biased outside-in posteriors indicate this sufficiency fails for some models.
  • ad hoc to paper The proportionality constant between the joint density and posterior is model-independent and cancels in the Bayes factor
    Explicitly assumed in Section 6 and later doubted in Section 8; this is the core assumption that makes the reported Bayes factors interpretable.
  • domain assumption Post-heated IGM (Ts >> TCMB) and zero inhomogeneous recombination (nrec = 0)
    Assumed in Section 2 for all lightcones; removes X-ray heating and recombination complexity from the toy models.
  • domain assumption The toy models share similar global reionization histories by construction
    The models are designed to match xHI evolution (Section 2.2), so model selection is based on morphology, not on reionization timing; this limits the generality of the test.
invented entities (2)
  • FZHinv outside-in toy model
    purpose: Inverted ionisation threshold to create an outside-in morphology for stress-testing the CNN
    Ad hoc model introduced in this paper; authors describe it as 'somewhat unphysical'. It has no observational prediction independent of the simulator.
  • MHR and MHRinv toy models
    purpose: Alternative reionization morphologies based on density-field ionisation thresholds; MHRinv is an inversion of MHR
    Adapted from Watkinson & Pritchard (2014) and Binnie & Pritchard (2019); they are test models for the method, not physical entities with independent evidence.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Likelihood-free Model Selection in Cosmic Reionization with Three-dimensional Tomographic 21 cm Lightcone Images." pith.science (2026). https://pith.science/paper/RNK6DT3L

@misc{pith2026250208152,
  author       = {Pith},
  title        = {Pith review of: Likelihood-free Model Selection in Cosmic Reionization with Three-dimensional Tomographic 21 cm Lightcone Images},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RNK6DT3L}},
  note         = {Machine review of arXiv:2502.08152}
}
read the original abstract

We explore likelihood-free (aka simulation-based) Bayesian model selection to quantify model comparison analyses of reionisation scenarios. We iteratively train the 3D Convolutional Neural Network (CNN) on four toy EoR models based on 21cmFAST simulations with contrasting morphology to obtain summaries of the 21 cm lightcone. Within the pyDelfi framework, we replaced the Emcee sampler with MultiNest to integrate learnt posteriors and produce the Bayesian Evidence. We comfortably distinguish the model used to produce the mock data set in all cases. However, we struggle to produce accurate posterior distributions for outside-in reionisation models. After a variety of cross-checks and alternate analyses we discuss the flexibility of summarising models that differ from precisely the intended network training conditions as this should be more widely scrutinised before CNN can reliably analyse observed data.

Figures

Figures reproduced from arXiv: 2502.08152 by the authors.

Figure 1
Figure 1. Brightness temperatures for the four toy EoR scenarios plotted z ∼ [7.5, 12] from the same density field. Namely, these are FZH, FZHinv, MHRinv, and MHR, in de￾scending order. FZH and MHRinv have an inside-out mor￾phology while FZHinv and MHR contrast with an outside-in morphology. A summary of these models is provided in Sec￾tion 2 [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. A summary of the 3D CNN architecture used in this work. For more detail about the purpose of each layer, please see Section 3. For a fuller description of the network architecture please see [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. We compare the output of Delfi-Nest and py￾Delfi on the cosmic shear model detailed in A19. In all cases, overlapping posterior distributions clearly show the agreement between the Emcee and MultiNest sampling techniques. rameters and specific prior details are shown in [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Both posteriors have been created with Delfi￾Nest, the only difference being the box size used when creat￾ing the training & mock data (128 Mpc2 in blue & 66 Mpc2 in red). Both show that Delfi-Nest can recover the fiducial parameters used to create the mock data. In re…
Figure 5
Figure 5. Figure 5: The Bayes factor results when reproducing each of the four mock data sets created by the four toy EoR models. Colour coding identifies the differing EoR morphologies as red for inside-out and blue for outside-in. In each plot, the model which produced the mock data is …
Figure 6
Figure 6. Figure 6: The parameter posteriors obtained when each model is fit against its own mock data set. These posteriors are a subset of the complete analysis shown in [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 7
Figure 7. Figure 7: Brightness temperature lightcones for the four toy EoR scenarios plotted z ∼ [7.5, 12] similar to [PITH_FULL_IMAGE:figures/full_fig_p020_7.png]
Figure 8
Figure 8. Figure 8: Each pair of plots shows the distribution of samples used to train each model’s 3D CNN (right) and the summary values obtained when summarising these lightcones with the 3D CNN (left). In Section B.4 we discuss increasing and decreasing the prior range by 10%, illustra…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Probabilistic cosmological inference on HI tomographic data

    astro-ph.IM 2025-07 conditional novelty 6.0 of 10

    A 3D CNN encoder plus a masked autoregressive flow recovers Ωm and σ8 from simulated HI tomographic data cubes at z=1 with R² ≥ 0.91 on test sets, with reduced but still useful accuracy out of distribution.

Reference graph

Works this paper leans on

140 extracted references · 10 canonical work pages · cited by 1 Pith paper

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter doi edition editor eprint howpublished institution journal key month number organization pages publisher school series title misctitle type volume year version url label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts ...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION format.url url empty "" new.block "" url * "" * if FUNCTION format.eprint eprint empty "" archivePrefix empty "" archivePrefix "arXiv" = new.block " " eprint * " " * new.block " " eprint * " " * if if if FUNCTION format.doi doi empty "" " " doi * " " * if FUNCTION format.pid doi empty eprint empty ur...

  3. [4]

    E., Alexander , P., et al

    Abdurashidova , Z., Aguirre , J. E., Alexander , P., et al. 2022, , 925, 221, 10.3847/1538-4357/ac1c78

    doi:10.3847/1538-4357/ac1c78 2022

  4. [5]

    2015, , 2015, 043, 10.1088/1475-7516/2015/08/043

    Akeret , J., Refregier , A., Amara , A., Seehars , S., & Hasner , C. 2015, , 2015, 043, 10.1088/1475-7516/2015/08/043

    doi:10.1088/1475-7516/2015/08/043 2015

  5. [6]

    2019, , 488, 4440, 10.1093/mnras/stz1960

    Alsing , J., Charnock , T., Feeney , S., & Wandelt , B. 2019, , 488, 4440, 10.1093/mnras/stz1960

    doi:10.1093/mnras/stz1960 2019

  6. [7]

    2018, , 476, L60, 10.1093/mnrasl/sly029

    Alsing , J., & Wandelt , B. 2018, , 476, L60, 10.1093/mnrasl/sly029

    doi:10.1093/mnrasl/sly029 2018

  7. [8]

    2018, , 477, 2874, 10.1093/mnras/sty819

    Alsing , J., Wandelt , B., & Feeney , S. 2018, , 477, 2874, 10.1093/mnras/sty819

    doi:10.1093/mnras/sty819 2018

  8. [9]

    Baldi, P., & Sadowski, P. J. 2013, NeurIPS, 26, https://proceedings.neurips.cc/paper_files/paper/2013/file/71f6278d140af599e06ad9bf1ba03cb0-Paper.pdf

    2013

Show all 140 references
  1. [10]

    2001, , 349, 125, 10.1016/S0370-1573(01)00019-9

    Barkana , R., & Loeb , A. 2001, , 349, 125, 10.1016/S0370-1573(01)00019-9

    2001 doi

  2. [11]

    Binnie , T., & Pritchard , J. R. 2019, , 487, 1160, 10.1093/mnras/stz1297

    2019 doi

  3. [12]

    Bishop , C. M. 1994, Technical Report, Mixture density networks, CiteSeer

    1994

  4. [13]

    M., & Nasrabadi , N

    Bishop , C. M., & Nasrabadi , N. M. 2007, JEI, 16, 049901, 10.1117/1.2819119

    2007 doi

  5. [14]

    Blum, M. G. B., Nunes, M. A., Prangle, D., & Sisson, S. A. 2013, Stat. Sci., 28, 189 , 10.1214/12-STS406

    2013 doi

  6. [15]

    R., Jaffe , A

    Bond , J. R., Jaffe , A. H., & Knox , L. 1998, , 57, 2117, 10.1103/PhysRevD.57.2117

    1998 doi

  7. [16]

    2000, , 533, 19, 10.1086/308625

    ---. 2000, , 533, 19, 10.1086/308625

    2000 doi

  8. [17]

    D., Rogers , A

    Bowman , J. D., Rogers , A. E. E., Monsalve , R. A., Mozdzen , T. J., & Mahesh , N. 2018 a , , 555, 67, 10.1038/nature25792

    2018 doi

  9. [18]

    2018 b , , 564, E35, 10.1038/s41586-018-0797-4

    ---. 2018 b , , 564, E35, 10.1038/s41586-018-0797-4

    2018 doi

  10. [19]

    2013, arXiv e-prints, 1301.6450, 10.48550/arXiv.1301.6450

    Cameron , E., & Pettitt , A. 2013, arXiv e-prints, 1301.6450, 10.48550/arXiv.1301.6450

    2013 doi arXiv

  11. [20]

    Charnock , T., Lavaux , G., & Wandelt , B. D. 2018, , 97, 083004, 10.1103/PhysRevD.97.083004

    2018 doi

  12. [21]

    2015, Keras, https://github.com/fchollet/keras, GitHub

    Chollet, F., et al. 2015, Keras, https://github.com/fchollet/keras, GitHub

    2015

  13. [22]

    2019, arXiv e-prints, 1910.06274, 1910.06274

    Cohen , A., Fialkov , A., Barkana , R., & Monsalve , R. 2019, arXiv e-prints, 1910.06274, 1910.06274

    2019 arXiv

  14. [23]

    E., Sainath, T

    Dahl, G. E., Sainath, T. N., & Hinton, G. E. 2013, ICASSP, 8609, 10.1109/ICASSP.2013.6639346

    2013

  15. [24]

    K., Jensen , H., Majumdar , S., et al

    Datta , K. K., Jensen , H., Majumdar , S., et al. 2014, , 442, 1491, 10.1093/mnras/stu927

    2014 doi

  16. [26]

    B., Hennawi , J

    Davies , F. B., Hennawi , J. F., Eilers , A.-C., & Luki \'c , Z. 2018, , 855, 106, 10.3847/1538-4357/aaaf70

    2018 doi

  17. [27]

    S., & Pacucci , F

    Dayal , P., Ferrara , A., Dunlop , J. S., & Pacucci , F. 2014, , 445, 2545, 10.1093/mnras/stu1848

    2014 doi

  18. [28]

    2020, arXiv e-prints, arXiv:2004.06709

    Deep Kaur , H., Gillet , N., & Mesinger , A. 2020, arXiv e-prints, arXiv:2004.06709. 2004.06709

    2020 arXiv

  19. [29]

    Ellis , R. S. 2014, arXiv e-prints, 1411.3330, 1411.3330

    2014 arXiv

  20. [30]

    A., et al

    Feng , Y., Di-Matteo , T., Croft , R. A., et al. 2016, , 455, 2778, 10.1093/mnras/stv2484

    2016 doi

  21. [32]

    P., & Bridges , M

    Feroz , F., Hobson , M. P., & Bridges , M. 2009, , 398, 1601, 10.1111/j.1365-2966.2009.14548.x

    2009

  22. [33]

    P., Cameron , E., & Pettitt , A

    Feroz , F., Hobson , M. P., Cameron , E., & Pettitt , A. N. 2019, DOAJ, 2, 10, 10.21105/astro.1306.2144

    2019 arXiv

  23. [34]

    2014, , 506, 197, 10.1038/nature12999

    Fialkov , A., Barkana , R., & Visbal , E. 2014, , 506, 197, 10.1038/nature12999

    2014 doi

  24. [35]

    Field , G. B. 1958, Proceedings of the IRE, 46, 240, 10.1109/JRPROC.1958.286741

    1958

  25. [36]

    D., Dav \'e , R., & Zackrisson , E

    Finlator , K., Keating , L., Oppenheimer , B. D., Dav \'e , R., & Zackrisson , E. 2018, , 480, 2628, 10.1093/mnras/sty1949

    2018 doi

  26. [37]

    W., Lang , D., & Goodman , J

    Foreman-Mackey , D., Hogg , D. W., Lang , D., & Goodman , J. 2013, , 125, 306, 10.1086/670067

    2013 doi

  27. [38]

    R., & Briggs , F

    Furlanetto , S. R., & Briggs , F. H. 2004, , 48, 1039, 10.1016/j.newar.2004.09.034

    2004 doi

  28. [39]

    R., Oh , S

    Furlanetto , S. R., Oh , S. P., & Briggs , F. H. 2006, , 433, 181, 10.1016/j.physrep.2006.08.002

    2006 doi

  29. [40]

    R., Zaldarriaga , M., & Hernquist , L

    Furlanetto , S. R., Zaldarriaga , M., & Hernquist , L. 2004, , 613, 1, 10.1086/423025

    2004 doi

  30. [41]

    2019, , 483, 5301, 10.1093/mnras/sty3414

    Garaldi , E., Compostella , M., & Porciani , C. 2019, , 483, 5301, 10.1093/mnras/sty3414

    2019 doi

  31. [42]

    2019, , 484, 282, 10.1093/mnras/stz010

    Gillet , N., Mesinger , A., Greig , B., Liu , A., & Ucci , G. 2019, , 484, 282, 10.1093/mnras/stz010

    2019 doi

  32. [43]

    2020, The Journal of Open Source Software, 5, 2363, 10.21105/joss.02363

    Giri , S., Mellema , G., & Jensen , H. 2020, The Journal of Open Source Software, 5, 2363, 10.21105/joss.02363

    2020 doi

  33. [44]

    K., Schneider , A., Maion , F., & Angulo , R

    Giri , S. K., Schneider , A., Maion , F., & Angulo , R. E. 2023, , 669, A6, 10.1051/0004-6361/202244986

    2023 doi

  34. [45]

    M., Hivon , E., Banday , A

    G \'o rski , K. M., Hivon , E., Banday , A. J., et al. 2005, , 622, 759, 10.1086/427976

    2005 doi

  35. [46]

    2015, , 449, 4246, 10.1093/mnras/stv571

    Greig , B., & Mesinger , A. 2015, , 449, 4246, 10.1093/mnras/stv571

    2015 doi

  36. [47]

    2017, , 472, 2651, 10.1093/mnras/stx2118

    ---. 2017, , 472, 2651, 10.1093/mnras/stx2118

    2017 doi

  37. [48]

    2018, , 477, 3217, 10.1093/mnras/sty796

    ---. 2018, , 477, 3217, 10.1093/mnras/sty796

    2018 doi

  38. [49]

    2019, , 483, 189, 10.1093/mnras/sty3102

    Hall , A., & Taylor , A. 2019, , 483, 189, 10.1093/mnras/sty3102

    2019 doi

  39. [50]

    R., Millman, K

    Harris, C. R., Millman, K. J., van der Walt, S., et al. 2020, CoRR, abs/2006.10256

    2020 arXiv

  40. [51]

    2019, , 483, 2524, 10.1093/mnras/sty3282

    Hassan , S., Liu , A., Kohn , S., & La Plante , P. 2019, , 483, 2524, 10.1093/mnras/sty3282

    2019 doi

  41. [52]

    Heavens, A., Alsing, J., & Jaffe, A. H. 2013, , 433, L6, 10.1093/mnrasl/slt045

    2013 doi

  42. [53]

    F., Jimenez , R., & Lahav , O

    Heavens , A. F., Jimenez , R., & Lahav , O. 2000, , 317, 965, 10.1046/j.1365-8711.2000.03692.x

    2000

  43. [54]

    F., Mootoovaloo , A., Trotta , R., & Sellentin , E

    Heavens , A. F., Mootoovaloo , A., Trotta , R., & Sellentin , E. 2023, , 2023, 048, 10.1088/1475-7516/2023/11/048

    2023 doi

  44. [55]

    2019, , 483, 2044, 10.1093/mnras/sty3090

    Higson , E., Handley , W., Hobson , M., & Lasenby , A. 2019, , 483, 2044, 10.1093/mnras/sty3090

    2019 doi

  45. [56]

    D., & Puchwein , E

    Hills , R., Kulkarni , G., Meerburg , P. D., & Puchwein , E. 2018, , 564, E32, 10.1038/s41586-018-0796-5

    2018 doi

  46. [57]

    2009, Bayesian Methods in Cosmology (CUP)

    Hobson , H., Jaffe , A., Liddle , A., Mukherjee , P., & D., P. 2009, Bayesian Methods in Cosmology (CUP)

    2009

  47. [58]

    1999, ApJ, 522, L21, 10.1086/312210

    Hu, W. 1999, ApJ, 522, L21, 10.1086/312210

    1999 doi

  48. [59]

    2001, Phys

    ---. 2001, Phys. Rev. D, 65, 023003, 10.1103/PhysRevD.65.023003

    2001 doi

  49. [60]

    Hunter, J. D. 2007, Computing in Science & Engineering, 9, 90, 10.1109/MCSE.2007.55

    2007 doi

  50. [61]

    2015, arXiv e-prints, 1502.03167, 10.48550/arXiv.1502.03167

    Ioffe , S., & Szegedy , C. 2015, arXiv e-prints, 1502.03167, 10.48550/arXiv.1502.03167

    2015 doi arXiv

  51. [62]

    Jaynes , E. T. 2003, Probability theory: The logic of science (CUP)

    2003

  52. [63]

    1992, , 388, 272, 10.1086/171151

    Kaiser , N. 1992, , 388, 272, 10.1086/171151

    1992 doi

  53. [64]

    1998, , 498, 26, 10.1086/305515

    ---. 1998, , 498, 26, 10.1086/305515

    1998 doi

  54. [65]

    A., Nabergoj , D., & Seljak , U

    Karamanis , M., Beutler , F., Peacock , J. A., Nabergoj , D., & Seljak , U. 2022 a , ASCL, ascl:2207.018

    2022

  55. [66]

    2022 b , JOSS, 7, 4634, 10.21105/joss.04634

    Karamanis , M., Nabergoj , D., Beutler , F., Peacock , J., & Seljak , U. 2022 b , JOSS, 7, 4634, 10.21105/joss.04634

    2022 doi

  56. [67]

    2018, CoRR, abs/1801.01450

    Kauderer - Abrams, E. 2018, CoRR, abs/1801.01450

    2018 arXiv

  57. [68]

    S., Liu , A., Parsons , A

    Kern , N. S., Liu , A., Parsons , A. R., Mesinger , A., & Greig , B. 2017, , 848, 23, 10.3847/1538-4357/aa8bb4

    2017 doi

  58. [69]

    2015, Rep

    Kilbinger , M. 2015, Rep. Prog. in Phys., 78, 086901, 10.1088/0034-4885/78/8/086901

    2015 doi

  59. [70]

    2015, PoS, AASKA14, 001, 10.22323/1.215.0001

    Koopmans, L., Pritchard, J., Mellema, G., et al. 2015, PoS, AASKA14, 001, 10.22323/1.215.0001

    2015 doi

  60. [71]

    2011, arXiv e-prints, 1110.3193, arXiv:1110.3193

    Laureijs , R., Amiaux , J., Arduini , S., et al. 2011, arXiv e-prints, 1110.3193, arXiv:1110.3193

    2011 arXiv

  61. [72]

    2019, GetDist: a Python package for analysing Monte Carlo samples

    Lewis, A. 2019, GetDist: a Python package for analysing Monte Carlo samples. 1910.13970

    2019 arXiv

  62. [73]

    Limber , D. N. 1954, , 119, 655, 10.1086/145870

    1954 doi

  63. [74]

    U., Dutta, R., Kaski, S., & Corander, J

    Lintusaari, J., Gutmann, M. U., Dutta, R., Kaski, S., & Corander, J. 2016, Syst. Biol., 66, e66, 10.1093/sysbio/syw077

    2016 doi

  64. [75]

    2013, The First Galaxies in the Universe (PUP)

    Loeb , A., & Furlanetto , S. 2013, The First Galaxies in the Universe (PUP)

    2013

  65. [76]

    Lueckmann , J.-M., Bassetto , G., Karaletsos , T., & Macke , J. H. 2018, arXiv e-prints, 1805.09294, 10.48550/arXiv.1805.09294

    2018 doi arXiv

  66. [77]

    J., Bassetto , G., et al

    Lueckmann , J.-M., Goncalves , P. J., Bassetto , G., et al. 2017, arXiv e-prints, 1711.01861, 10.48550/arXiv.1711.01861

    2017 doi arXiv

  67. [78]

    2012, arXiv e-prints, 1205.2629, 10.48550/arXiv.1205.2629

    Lyu , S. 2012, arXiv e-prints, 1205.2629, 10.48550/arXiv.1205.2629

    2012 doi arXiv

  68. [79]

    1998, NeurIPS, 11, https://proceedings.neurips.cc/paper_files/paper/1998/file/9327969053c0068dd9e07c529866b94d-Paper.pdf

    Magdon-Ismail, M., & Atiya, A. 1998, NeurIPS, 11, https://proceedings.neurips.cc/paper_files/paper/1998/file/9327969053c0068dd9e07c529866b94d-Paper.pdf

    1998

  69. [80]

    A., Mu \ n oz , J

    Mason , C. A., Mu \ n oz , J. B., Greig , B., Mesinger , A., & Park , J. 2023, , 524, 4711, 10.1093/mnras/stad2145

    2023 doi

  70. [81]

    D., Beckman, R

    McKay, M. D., Beckman, R. J., & Conover, W. J. 1979, Technometrics, 21, 239, http://www.jstor.org/stable/1268522

    1979

  71. [82]

    Melchior , P., & Goulding , A. D. 2018, A& C, 25, 183, 10.1016/j.ascom.2018.09.013

    2018 doi

  72. [83]

    T., Alvarez , M

    Mellema , G., Iliev , I. T., Alvarez , M. A., & Shapiro , P. R. 2006, , 11, 374, 10.1016/j.newast.2005.09.004

    2006 doi

  73. [84]

    Mellema , G., Koopmans , L. V. E., Abdalla , F. A., et al. 2013, Exp. Astron., 36, 235, 10.1007/s10686-013-9334-5

    2013 doi

  74. [85]

    G., Mevius , M., Koopmans , L

    Mertens , F. G., Mevius , M., Koopmans , L. V. E., et al. 2020, , 493, 1662, 10.1093/mnras/staa327

    2020 doi

  75. [86]

    2007, , 669, 663, 10.1086/521806

    Mesinger , A., & Furlanetto , S. 2007, , 669, 663, 10.1086/521806

    2007 doi

  76. [88]

    Miralda-Escud \'e , J., Haehnelt , M., & Rees , M. J. 2000, , 530, 1, 10.1086/308330

    2000 doi

  77. [89]

    2015, , 449, L41, 10.1093/mnrasl/slv015

    Mondal , R., Bharadwaj , S., Majumdar , S., Bera , A., & Acharyya , A. 2015, , 449, L41, 10.1093/mnrasl/slv015

    2015 doi

  78. [90]

    A., Bye , C

    Monsalve , R. A., Bye , C. H., Sievers , J. L., et al. 2023, arXiv e-prints, 2310.07741, 10.48550/arXiv.2310.07741

    2023 doi arXiv

  79. [91]

    2020, JOSS, 5, 2582, 10.21105/joss.02582

    Murray , S., & et al. 2020, JOSS, 5, 2582, 10.21105/joss.02582

    2020 doi

  80. [92]

    2022, , 511, 3446, 10.1093/mnras/stac218

    Neutsch , S., Heneka , C., & Br \"u ggen , M. 2022, , 511, 3446, 10.1093/mnras/stac218

    2022 doi

  81. [93]

    R., et al

    Ocvirk , P., Gillet , N., Shapiro , P. R., et al. 2016, , 463, 1462, 10.1093/mnras/stw2036

    2016 doi

  82. [94]

    G., et al

    Ocvirk , P., Aubert , D., Sorce , J. G., et al. 2018, arXiv e-prints, 1811.11192, 1811.11192

    2018 arXiv

  83. [95]

    2020, , 498, 373, 10.1093/mnras/staa2118

    Pagano , M., & Liu , A. 2020, , 498, 373, 10.1093/mnras/staa2118

    2020 doi

  84. [96]

    2019, arXiv e-prints, 1910.13233, 10.48550/arXiv.1910.13233

    Papamakarios , G. 2019, arXiv e-prints, 1910.13233, 10.48550/arXiv.1910.13233

    2019 doi arXiv

  85. [97]

    2016, arXiv e-prints, 1605.06376, 1605.06376

    Papamakarios , G., & Murray , I. 2016, arXiv e-prints, 1605.06376, 1605.06376

    2016 arXiv

  86. [98]

    2017, arxiv e-prints, 1705.07057, 1705.07057

    Papamakarios , G., Pavlakou , T., & Murray , I. 2017, arxiv e-prints, 1705.07057, 1705.07057

    2017 arXiv

  87. [99]

    2019, , 484, 933, 10.1093/mnras/stz032

    Park , J., Mesinger , A., Greig , B., & Gillet , N. 2019, , 484, 933, 10.1093/mnras/stz032

    2019 doi

  88. [100]

    Pattison , J. H. N., Anstey , D. J., & de Lera Acedo , E. 2023, arXiv e-prints, 2307.02908, 10.48550/arXiv.2307.02908

    2023 doi arXiv

  89. [101]

    2011, Journal of Machine Learning Research, 12, 2825

    Pedregosa, F., Varoquaux, G., Gramfort, A., et al. 2011, Journal of Machine Learning Research, 12, 2825

    2011

  90. [102]

    Planck Collaboration , Ade , P. A. R., Aghanim , N., et al. 2016, , 594, A13, 10.1051/0004-6361/201525830

    2016 doi

  91. [103]

    2020, , 641, A6, 10.1051/0004-6361/201833910

    Planck Collaboration , Aghanim , N., Akrami , Y., et al. 2020, , 641, A6, 10.1051/0004-6361/201833910

    2020 doi

  92. [104]

    2022, , 509, 3852, 10.1093/mnras/stab3215

    Prelogovi \'c , D., Mesinger , A., Murray , S., Fiameni , G., & Gillet , N. 2022, , 509, 3852, 10.1093/mnras/stab3215

    2022 doi

  93. [105]

    R., & Furlanetto , S

    Pritchard , J. R., & Furlanetto , S. R. 2006, , 367, 1057, 10.1111/j.1365-2966.2006.10028.x

    2006

  94. [106]

    2007, , 376, 1680, 10.1111/j.1365-2966.2007.11519.x

    ---. 2007, , 376, 1680, 10.1111/j.1365-2966.2007.11519.x

    2007

  95. [107]

    R., & Loeb , A

    Pritchard , J. R., & Loeb , A. 2012, Rep. Prog. in Phys., 75, 086901, 10.1088/0034-4885/75/8/086901

    2012 doi

  96. [108]

    G., Ferramacho , L., Silva , M

    Santos , M. G., Ferramacho , L., Silva , M. B., Amblard , A., & Cooray , A. 2010, , 406, 2421, 10.1111/j.1365-2966.2010.16898.x

    2010

  97. [109]

    2018, NeurIPS, 2488

    Santurkar, S., Tsipras, D., Ilyas, A., & Madry, A. 2018, NeurIPS, 2488. http://dblp.uni-trier.de/db/conf/nips/nips2018.html#SanturkarTIM18

    2018

  98. [110]

    J., & Pritchard , J

    Schmit , C. J., & Pritchard , J. R. 2018, , 475, 1213, 10.1093/mnras/stx3292

    2018 doi

  99. [111]

    Sellentin , E., & Heavens , A. F. 2016, , 456, L132, 10.1093/mnrasl/slv190

    2016 doi

  100. [112]

    Semih Kayhan, O., & van Gemert, J. C. 2020, in 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 14262--14273, 10.1109/CVPR42600.2020.01428

    2020

  101. [113]

    2017, , 468, 3869, 10.1093/mnras/stx734

    Shimabukuro , H., & Semelin , B. 2017, , 468, 3869, 10.1093/mnras/stx734

    2017 doi

  102. [114]

    2018, , 858, 54, 10.3847/1538-4357/aabae1

    Singh , S., Subrahmanyan , R., Udaya Shankar , N., et al. 2018, , 858, 54, 10.3847/1538-4357/aabae1

    2018 doi

  103. [115]

    2004, AIP Conf Proc, 735, 395, 10.1063/1.1835238

    Skilling , J. 2004, AIP Conf Proc, 735, 395, 10.1063/1.1835238

    2004 doi

  104. [116]

    2014, , 440, 1662, 10.1093/mnras/stu377

    Sobacchi , E., & Mesinger , A. 2014, , 440, 1662, 10.1093/mnras/stu377

    2014 doi

  105. [117]

    Speagle , J. S. 2020, , 493, 3132, 10.1093/mnras/staa278

    2020 doi

  106. [118]

    2014, JMLR, 15, 1929, http://jmlr.org/papers/v15/srivastava14a.html

    Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., & Salakhutdinov, R. 2014, JMLR, 15, 1929, http://jmlr.org/papers/v15/srivastava14a.html

    2014

  107. [119]

    2004, , 348, 897, 10.1111/j.1365-2966.2004.07410.x

    Takada , M., & Jain , B. 2004, , 348, 897, 10.1111/j.1365-2966.2004.07410.x

    2004

  108. [120]

    2013, , 432, 1928, 10.1093/mnras/stt270

    Taylor , A., Joachimi , B., & Kitching , T. 2013, , 432, 1928, 10.1093/mnras/stt270

    2013 doi

  109. [121]

    1997, , 55, 5895, 10.1103/PhysRevD.55.5895

    Tegmark , M. 1997, , 55, 5895, 10.1103/PhysRevD.55.5895

    1997 doi

  110. [122]

    N., & Heavens , A

    Tegmark , M., Taylor , A. N., & Heavens , A. F. 1997, , 480, 22, 10.1086/303939

    1997 doi

  111. [123]

    2012, COURSERA: Neural networks for machine learning, 4, 26

    Tieleman, T., & Hinton, G. 2012, COURSERA: Neural networks for machine learning, 4, 26

    2012

  112. [124]

    Trott , C. M. 2016, , 461, 126, 10.1093/mnras/stw1310

    2016 doi

  113. [125]

    2008, Contemporary Physics, 49, 71, 10.1080/00107510802066753

    Trotta , R. 2008, Contemporary Physics, 49, 71, 10.1080/00107510802066753

    2008 doi

  114. [126]

    Van Rossum, G., & Drake, F. L. 2009, Python 3 Reference Manual (Scotts Valley, CA: CreateSpace)

    2009

  115. [127]

    E., et al

    Virtanen, P., Gommers, R., Oliphant, T. E., et al. 2019, CoRR, abs/1907.10121

    2019 arXiv

  116. [128]

    A., Majumdar , S., Pritchard , J

    Watkinson , C. A., Majumdar , S., Pritchard , J. R., & Mondal , R. 2017, , 472, 2436, 10.1093/mnras/stx2130

    2017 doi

  117. [129]

    A., & Pritchard , J

    Watkinson , C. A., & Pritchard , J. R. 2014, , 443, 3090, 10.1093/mnras/stu1384

    2014 doi

  118. [130]

    A., Trott , C

    Watkinson , C. A., Trott , C. M., & Hothi , I. 2020, arXiv e-prints, 2002.05992, 2002.05992

    2020 arXiv

  119. [131]

    A., & Storath, M

    Weiler, M., Hamprecht, F. A., & Storath, M. 2018, in 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 849--858, 10.1109/CVPR.2018.00095

    2018

  120. [132]

    Wilkinson, R. D. 2013, Stat. Appl. Genet. Mol. Biol., 12, 129, doi:10.1515/sagmb-2013-0010

    2013 doi

  121. [133]

    Wise , J. H. 2019, arXiv e-prints, 1907.06653, arXiv:1907.06653

    2019 arXiv

  122. [134]

    H., Abel , T., Turk , M

    Wise , J. H., Abel , T., Turk , M. J., Norman , M. L., & Smith , B. D. 2012 a , , 427, 311, 10.1111/j.1365-2966.2012.21809.x

    2012

  123. [135]

    H., Demchenko , V

    Wise , J. H., Demchenko , V. G., Halicek , M. T., et al. 2014, , 442, 2560, 10.1093/mnras/stu979

    2014 doi

  124. [136]

    H., Turk , M

    Wise , J. H., Turk , M. J., Norman , M. L., & Abel , T. 2012 b , , 745, 50, 10.1088/0004-637X/745/1/50

    2012 doi

  125. [137]

    Yeh , J. Y. C., Smith , A., Kannan , R., et al. 2023, , 10.1093/mnras/stad210

    2023 doi

  126. [138]

    Yoshiura , S., Pindor , B., Line , J. L. B., et al. 2021, , 505, 4775, 10.1093/mnras/stab1560

    2021 doi

  127. [139]

    2016, , 93, 083525, 10.1103/PhysRevD.93.083525

    Zablocki , A., & Dodelson , S. 2016, , 93, 083525, 10.1103/PhysRevD.93.083525

    2016 doi

  128. [140]

    2011, , 414, 727, 10.1111/j.1365-2966.2011.18439.x

    Zahn , O., Mesinger , A., McQuinn , M., et al. 2011, , 414, 727, 10.1111/j.1365-2966.2011.18439.x

    2011

  129. [141]

    Zel'Dovich , Y. B. 1970, , 500, 13

    1970

  130. [142]

    Zhao , X., Mao , Y., Cheng , C., & Wandelt , B. D. 2022 a , , 926, 151, 10.3847/1538-4357/ac457d

    2022 doi

  131. [143]

    Zhao , X., Mao , Y., & Wandelt , B. D. 2022 b , , 933, 236, 10.3847/1538-4357/ac778e

    2022 doi

  132. [144]

    Zhao, X., Mao, Y., Zuo, S., & Wandelt, B. D. 2023, arXiv e-prints, 2310.17602v1, 2310.17602v1

    2023 arXiv

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.